Batch calculated-channel solve into temp tables, mirroring aggregations/events

[tombonfert]

What

Calculated channels are now solved in batches, following the same pattern
already used for aggregations and events. The batching is driven from the Report
class: the channels are partitioned by query_engine.batch_size, each batch is
solved and persisted as a Unity Catalog temp table
(__impulse_temp_{run_id}_{batch_idx}), and the batches are unioned into the final
calculated_channel_fact.

Why

Previously all calculated channels were solved in a single
query.select(*channels).solve_calculated_channels(...) call, unlike aggregations
and events which batch by batch_size and persist intermediate results as temp
tables. Large channel sets therefore built one big lazy plan with no intermediate
materialization. This change gives calculated channels the same bounded,
batch-at-a-time execution and a consistent orchestration shape across all entity
types.

How

• New solve_calculated_channels_batched (report_utils.py): the narrow,
  row-append counterpart to solve_expressions_batched. It reuses build_batches
  (selector-aware, batch_size = max unique selectors per batch), solves each batch
  via query.select(*batch).solve_calculated_channels(...), persists each as a temp
  Delta table (or a Spark temp view when sinkless), and combines them with
  unionByName. This is the one deliberate difference from the wide solver, whose
  batches are one-row-per-container and combined with a container_id join.
• Batching moved to Report: new Report._solve_calculated_channels_batched
  (parallel to _solve_expressions_batched); determine_report collects the
  changed/unchanged channel expressions, batch-solves each into a narrow
  solved_df, and dispatches those.
• dispatch_calculated_channels now takes solved_df (dropping
  query/solver), mirroring dispatch_aggregations.
• CalculatedChannel.determine_calculated_channels no longer solves; it shapes
  the already-solved solved_df (filters by channel_id, projects to the fact
  schema), like determine_aggregations.
• Temp tables share the existing __impulse_temp_* prefix, so the start-of-run
  cleanup and the unity_sink.cleanup_temp_tables flag cover them automatically. No
  new config; reuses query_engine.batch_size. The add_calculated_channel API is
  unchanged.

Tests

• Unit: TestSolveCalculatedChannelsBatched (sinkless temp view, sink Delta table,
  unionByName not join, per-batch select, unique run_id) and the reworked
  TestDetermineCalculatedChannels (shaping: filter by channel_id + project).
• Integration: test_batched_calculated_channels_union runs batch_size=1 with two
  distinct-selector channels and asserts both land in the union plus two temp tables
  are created; the RAW-mode test is updated to the new solve path.
• Regression: the full report_utils unit suite and the batched-pipeline
  integration test confirm the shared batching/cleanup path still works for
  aggregations and events.

Docs

Updated the channel reference, the impulse-channels skill, and the batch_size
descriptions in the config doc/skill; regenerated the pydoc-markdown API reference.

Test Plan

• Unit tests added/updated
• Manual testing completed
• Documentation updated (if applicable)

Checklist

• Code follows project style guidelines
• Self-review completed
• No new linter warnings introduced

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.89%. Comparing base (d0418d9) to head (3529a71).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #80      +/-   ##
==========================================
+ Coverage   88.83%   88.89%   +0.05%     
==========================================
  Files          61       61              
  Lines        5170     5194      +24     
  Branches      621      625       +4     
==========================================
+ Hits         4593     4617      +24     
  Misses        466      466              
  Partials      111      111              

Flag	Coverage Δ
query_engine	84.57% <ø> (ø)
reporting	94.22% <100.00%> (+0.06%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...c/impulse_reporting/channels/calculated_channel.py	100.00% <100.00%> (ø)
src/impulse_reporting/core/report.py	92.06% <100.00%> (+0.15%)	⬆️
src/impulse_reporting/core/report_utils.py	97.41% <100.00%> (+0.21%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.